Model routing is deciding which class of AI model runs each job, by rule rather than by preference. High-volume work goes to fast, small models; the most capable class is held for work that genuinely needs it; and nobody using the product has to choose.
Why it matters
New models launch every few weeks, each with a new name and a new claim. For anyone trying to run a business on top of them, keeping up is a job nobody asked for.
Picking one model for everything fails in both directions. The cheap choice cannot carry the hard work, and the expensive choice burns budget on jobs a small model would have finished correctly.
So the decision belongs in a table rather than a settings screen. Match each job to the class of model it needs, and change the table when something better ships.
How Jinn treats it
Every job in the Jinn suite is routed to the class of model it needs, fast and small where volume matters, the most capable class only where the work truly demands it, and that routing stays current as vendors ship new families.
It is one table rather than a preference, which is why Jinn names classes of model instead of brand-name models: the rules are written to outlive each launch. High-volume jobs are pinned to fast, small models by rule, so a premium plan buys more coverage rather than an oversized model on work that never needed one.
One routing table maps each job to a model class, and the whole suite moves when it updates. See it
Related terms
You never have to pick a model.
Each job goes to the class it needs, by rule, and the table moves when the market does. See how routing works.